Different weighting methods in business tendency survey indicators in Swiss manufacturing industry

Abstract

In business cycle analysis the development of inventories still plays a crucial role. The strong movements over time have a large effect particularly on the growth rate of GDP. Quantitative data on inventories are generally of rather low quality. As a complement to quantitative statistics, business tendency surveys (BTS) offer actual data on stocks of different categories and on order books (often called negative inventories) to estimate actual values. However, the qualitative data do not show up in the empirical analysis. One reason could be that the aggregation process in BTS of these two items is often not adequate. Three alternative weighting methods were applied at firm level: no weighting, weighting with the number of employees, weighting with the number of employees plus the ratio of order books/stock to sales. These were compared with the current weighting method which includes stratification and branch weights by value added. The four indicators for each of the BTS questions had a statistically different variance, mean or distribution in most cases. The comparison of these four versions of weighting with the reference series – growth rate of order books and of stocks of finished products – produced quite different results. For the growth rate of order books, the best fit was with the non-weighted responses of the firms to the question on changes of order books. The match with the growth rate of stocks of finished products was generally lower. The best fit was again with the non-weighted responses on the change question on stocks of finished products.